Lindahl Lecture 3: Housing, Transportation Technology and City Governments Edward L. Glaeser Harvard University
Dec 25, 2015
Lindahl Lecture 3: Housing, Transportation Technology and
City GovernmentsEdward L. Glaeser
Harvard University
Structure of Lecture
• Transportation Technologies and Cities– Urban Poverty – Sprawl
• Housing Demand and Supply• Government Policies towards Housing
– Rent Control– Subsidized Homeownership– General Aside on Social Capital
• Cities and Governments
Why do the poor live in central cities?
• Poverty rate in central cities is 18%; in suburbs it is 8%
• In old metropolitan areas, poverty frises and then falls with distance from CBD
• In newer metro areas, poverty just declines with distance from CBD
• New migrants to cities are just as poor as old residents– selection not treatment
The AMM Model
• With two groups the willingness to pay for proximity per acre determines who lives closer to the city center
• The key willingness to pay comes is P’(d) from P(d)a+dt= total costs, so willingness to pay is for proximity is -P’(d)=t/a
• This means that the poor live in cities if they have higher commuting costs or less demand for land
The Model Graphically
House Price
Distance
Whoever has a steeper curveLives near the center
The Elasticity Condition
• With two groups, the question is whether tr / ar >tp / ap or
or or
pr
pr
p
p
pr
pr
p
p
p
pr
p
pr
p
r
p
r
yy
aa
a
y
yy
tt
t
y
a
aa
t
tt
a
a
t
t
Poor in Cities continued
With income as a continuum, the key question is whether the income elasticity of demand for land is greater or less than the income elasticity of commuting costs.
What is a reasonable benchmark for the income elasticity of commuting costs? One mode– probably somewhat under 1.
The Income Elasticity of Demand for Land
• If you just look at people live in single family detached houses: .1
• This rises to .3 if you instrument with education for income (perm. Income)
• But apartments are the critical issue, and then you need to assign land to an apartment.
• Our best estimate is .3-.4.
Why do the poor live in cities?
• The role of transport modes– if the rich drive and the poor take transit the puzzle can be resolved.
• Even though the rich pay more for driving the marginal cost of distance is less
• Cars move (on average) at 30 mph and have a minute fixed cost
• Buses move at under 20 on average• The income elasticity of auto ownership is high
Evidence on the Role of Public Transportation
• Cross-section– people who live close to public transportation are poorer holding distance to CBD fixed– Rail in Boston, Portland, Washington– Buses in LA– Subways in NYC (outer buroughs)
• Panel– when tracts get new access to trains, the poverty rate rises
Evidence continued
• In areas where everyone drives, the rich live closer to the city
• The existence of subways creates a zone where the rich take public transport and dense cities the rich walk in the center– In these cities, the relationship between
income and distance is not monotonic
• Statistically, these subways change the urban form
Why the 20th century transformation?
• The move to sun and sprawl both reflect the same phenomenon.
• Transportation costs have fallen.
• Consumer cities– not producer cities.
• Car cities, not walking or PT cities.
Reduction of the Costs of Moving Goods
• Railroads, Trucking, Highways have radically reduced transport costs.
• Manufacturing no longer locates near its suppliers/consumers.
• Manufacturing has suburbanized and left cities (and the US) altogether.
• Boston was typical– not unique.
Declining Transport Costs: RailDollars per ton-mile (real)
Year1890 2000
0.023
0.185
Dollars per ton-mile (real)
Year1890 2000
0.023
0.185
Declining Costs: More Modes
Transportation share of GDP Without air
1850 1900 1950 2000
0.02
0.04
0.06
0.08
0.10
Year
Share of GDPin transport
Transportation share of GDP Without air
1850 1900 1950 2000
0.02
0.04
0.06
0.08
0.10
0.02
0.04
0.06
0.08
0.10
Year
Share of GDPin transport
As Transport Costs Fell– Manufacturing Left Cities
• First is suburbanized– Manufacturing firms are big users of space – There is a strong tendency of these firms to locate for
from the city center. • Then it left high density counties
• And it left the U.S.
• There is no reason to think that this is inefficient or bad.
Manufacturing and Density
Log of people per square mile
Manufacturing share
-1.84 11.08
-0.02
0.54
Log of people per square mile
Manufacturing share
-1.84 11.08
-0.02
0.54
Manufacturing and Decline 1920-1980
Figure 9: Manufacturing and Urban DeclineManufacturing Employment Share
Population Grow th 20-80 .
0 20 40
-1
0
1
2
3
SAN DIEG
LOS ANGE
DULUTH,
SIOUX CI
NEW ORLE SALT LAKDES MOIN
MANCHEST
TACOMA, SPOKANE,
KNOXVILL
FLINT, M
WATERBUR
FALL RIV
BIRMINGH
WICHITA,
HOUSTON,
SPRINGFI
PORTLAND
DENVER,
ST. PAUL
SAN ANTO
SOUTH BEOAKLAND,
ERIE, PA
WORCESTE
YONKERS,
OMAHA, N
YOUNGSTO
OKLAHOMA
SEATTLE,
FORT WAY
KANSAS C
CINCINNAUTICA, N
ST. JOSE
JACKSONV
ALBANY, NEW BEDF
KANSAS C
EL PASO,
FORT WORNASHVILL
TROY, NY
MEMPHIS,
DALLAS,
CANTON,
SCRANTON
RICHMOND
INDIANAP
WASHINGT ALLENTOW
ATLANTA,
MINNEAPO
GRAND RA
PEORIA, TOLEDO,
LOWELL,
HARTFORD
NEW HAVE
AKRON, O
TULSA, O
BALTIMORSYRACUSE
DAYTON, EVANSVIL
BRIDGEPOELIZABET
ROCHESTE
LOUISVIL
COLUMBUS
READING,SCHENECT
SAVANNAH
SAN FRAN
ST. LOUI
DETROIT,
HARRISBU BUFFALO, PROVIDEN
CHICAGO,
LAWRENCECLEVELANWILKES-B
PHILADEL
PITTSBUR CAMDEN,
NORFOLK,
WILMINGT
PATERSON
TRENTON,BOSTON, CAMBRIDG
NEWARK,
MILWAUKENEW YORK
BAYONNE,JERSEY C
SOMERVIL
The Rise of Car Cities
• First, there was flight to the suburbs.
• Then the jobs left too– Now more than 75 percent of Chicago’s jobs are outside the classic downtown.
• Firms followed people (again consumer cities).
• Movement both within MSAs to edge cities and across MSAs to car friendly places.
Density and Decline: 1920-1980
Figure 8: Density and City Growth 1920-1980dens20
Population Grow th 20-80 Fitted values
947.754 23869.5
-.545373
2.46159
NEW YORKCHICAGO,
PHILADEL
DETROIT,
CLEVELAN
ST. LOUI
BOSTON,
BALTIMOR
PITTSBUR
LOS ANGE
BUFFALO,
SAN FRAN MILWAUKEWASHINGT
NEWARK,
CINCINNA
NEW ORLE
MINNEAPO
KANSAS CSEATTLE,
INDIANAP
JERSEY CROCHESTE
PORTLAND
DENVER,
TOLEDO,
PROVIDEN
COLUMBUS
LOUISVILST. PAUL
OAKLAND,
AKRON, O
ATLANTA,
OMAHA, N
WORCESTE
BIRMINGH
SYRACUSE
RICHMOND
NEW HAVE
MEMPHIS,
SAN ANTO
DALLAS,
DAYTON,
BRIDGEPO
HOUSTON,
HARTFORD
SCRANTON
GRAND RA
PATERSON
YOUNGSTO
SPRINGFI
DES MOIN
NEW BEDFFALL RIV TRENTON,
NASHVILL
SALT LAK
CAMDEN,
NORFOLK,
ALBANY, LOWELL,
WILMINGT
CAMBRIDG
READING,
FORT WOR
SPOKANE, KANSAS C
YONKERS,
DULUTH,
TACOMA,
ELIZABET
LAWRENCE
UTICA, N
ERIE, PA
SOMERVIL
WATERBUR
FLINT, M
JACKSONV
OKLAHOMA
SCHENECT
CANTON,
FORT WAY
EVANSVILSAVANNAH
MANCHEST
ST. JOSE
KNOXVILL
EL PASO,
BAYONNE,
PEORIA,
HARRISBU
SAN DIEG
WILKES-B
ALLENTOW
WICHITA,
TULSA, O
TROY, NY
SIOUX CI
SOUTH BE
Figure 8: Density and City Growth 1920-1980dens20
Population Grow th 20-80 Fitted values
947.754 23869.5
-.545373
2.46159
NEW YORKCHICAGO,
PHILADEL
DETROIT,
CLEVELAN
ST. LOUI
BOSTON,
BALTIMOR
PITTSBUR
LOS ANGE
BUFFALO,
SAN FRAN MILWAUKEWASHINGT
NEWARK,
CINCINNA
NEW ORLE
MINNEAPO
KANSAS CSEATTLE,
INDIANAP
JERSEY CROCHESTE
PORTLAND
DENVER,
TOLEDO,
PROVIDEN
COLUMBUS
LOUISVILST. PAUL
OAKLAND,
AKRON, O
ATLANTA,
OMAHA, N
WORCESTE
BIRMINGH
SYRACUSE
RICHMOND
NEW HAVE
MEMPHIS,
SAN ANTO
DALLAS,
DAYTON,
BRIDGEPO
HOUSTON,
HARTFORD
SCRANTON
GRAND RA
PATERSON
YOUNGSTO
SPRINGFI
DES MOIN
NEW BEDFFALL RIV TRENTON,
NASHVILL
SALT LAK
CAMDEN,
NORFOLK,
ALBANY, LOWELL,
WILMINGT
CAMBRIDG
READING,
FORT WOR
SPOKANE, KANSAS C
YONKERS,
DULUTH,
TACOMA,
ELIZABET
LAWRENCE
UTICA, N
ERIE, PA
SOMERVIL
WATERBUR
FLINT, M
JACKSONV
OKLAHOMA
SCHENECT
CANTON,
FORT WAY
EVANSVILSAVANNAH
MANCHEST
ST. JOSE
KNOXVILL
EL PASO,
BAYONNE,
PEORIA,
HARRISBU
SAN DIEG
WILKES-B
ALLENTOW
WICHITA,
TULSA, O
TROY, NY
SIOUX CI
SOUTH BE
Car Cities in the 1990s: Is there a New Urbanism?
avg. vehicles available per hhol
growth Fitted values
.6 2.3
-.129964
.852277
Facts about Sprawl
• In most American cities, more than 80 percent of people live more than 3 miles from the CBD
• More than 75 percent of workers work outside that ring
• In cities with decentralized employment, rents don’t rise much with distance and commute times don’t rise
Sprawl Cities are Car Cities
• 92 percent of trips are by car• Even 77 percent of trips under a mile are
by car• Places with more African-Americans in the
center have more dispersion– but the differences are small
• Across countries, using gas prices instrumented for by legal origin predicts sprawl
Is Sprawl Bad?
• Pollution– potentially serious for global warming, but most other problems have been taken car of by technology
• Little land area is actually used in the U.S.• Congestion– sure, but it is not obvious that
congestion rises with sprawl– commute times actually fall
• Commute time by car is 23 minutes on average 47 minutes by public transportation
• The big plus is housing size which reached over 2000 square feet in the last few years
Cars and Driving Times
% using public transportation 19
average travel time to work (mi Fitted values
.2 53.4
17
36.8513
Albuquer
Anaheim
Arlingto
Atlanta
Austin c
Baltimor
Birmingh
Boston c
Buffalo
Charlott
Chicago
Cincinna
Clevelan
Colorado
Columbus
Corpus C
Dallas c
Denver c
Detroit
El Paso
Fort Wor
Fresno c
Honolulu
Houston
IndianapJacksonv
Kansas CLas Vega
Long Bea
Los Ange
Louisvil
Memphis
Mesa cit
Miami ci
MilwaukeMinneapo
Nashvill
New Orle
New York
Newark c
Norfolk
Oakland
Oklahoma
Omaha ci
Philadel
Phoenix
PittsburPortland
Sacramen
San Anto
San Dieg
San Fran
San Jose
Santa An
Seattle St. Loui
St. PaulTampa ci
Toledo c
Tucson c
Tulsa ci
Virginia
Washingt
Wichita
The Europeans and their Trains
• Fact # 1: In rich European cities, people now drive just like in the U.S.
• Fact # 2: In many cities where people rarely drive, commute times are very high:– Moscow 10% drive, 62 minute commute.– Athens 36% drive, 53 minute commute.– Paris 60% drive, 35 minute commute. – US average is 24 minutes.
Cars and Travel Time Internationally
% wrk trips by private car
mean travel time to work (mins) Fitted values
2 81
14
78
Yerevan
Melbourn
Gaborone
Brasilia
Curitiba
Recife
Rio de J
Bandar S
Sofia
Toronto
Santiago
Brazzavi Zagreb
Camaguey
La Haban
Prague
Copenhag Djibouti BordeauxDunkerqu
Paris
RennesStrasbou
Tbilisi
Athens
Budapest
Amman
Almaty
RigaVilnius
Birkirka
ChisinauAmsterda
Buchares
Kostroma
Moscow
Nizhny N
Novgorod
Ryazan
Sao Tome
BratislaBratisla
Koper
Ljubljan
Maribor
Stockhol
Dubai
Hertford
New YorkBelgrade
Nis
Novi Sad
Housing Demand and Supply
• Traditionally urban literature has focused on housing demand– using housing price hedonics to back out demand for place
• New literature focuses more on supply, in part because supply drives city growth
• In part because recent regulatory changes are incredibly important and underexplored
Why Supply MattersLo
g H
ous
ing
Uni
ts, 2
000
Figure 1: Housing Units and Population Levels, 2000Log Population, 2000
log total housing units 2000 Fitted values
10.0819 15.896
9.25387
14.9789
JohnstowAnnistonNew LondPoughkeeWilliams
WheelingJamestowHouma ciParkersbTexarkan
FlorenceJackson HagerstoBremerto
State Co
Wausau cFitchburLima citMuskegonAtlanticYork citMiddletoBurlingt
CharlottJoplin cAlexandrBinghamtDanville
New Brun
HarrisbuMansfielAltoona Biloxi cHuntingtElkhart SarasotaMonroe cBattle CCharlestDecatur Galvesto
Santa CrJohnson Niagara GreenvilPensacolLancasteLafayettDothan cFayettevJanesvilTerre HaAndersonUtica ciHamiltonEau ClaiSaginaw Santa FeDaytona PortlandBloomingLynchburBryan ci
BellinghMuncie cChampaigLorain cWaterlooAshevill
BloomingAppletonMelbournLake ChaYakima cLawrenceWilmingtPawtuckeLongviewLas CrucDanbury
WilmingtScrantonGreeley Albany cKalamazo
Santa Ma
TuscalooLakelandFort SmiCanton cReading Racine cDecatur YoungstoWest PalNorwalk Tyler ciColumbiaSioux CiTrenton RochesteKilleen Duluth cSan AngeBillingsKenosha
Fargo ciOdessa cFall RivSanta BaLawton cNew BedfBrocktonBoulder
Roanoke Midland GainesviAlbany cCharlestMacon ciDavenporLivonia PortsmouPueblo cBerkeleyGary citClarksvi
Erie citWichita
Provo ci
Lowell cJoliet cMcAllen AllentowManchestWaterburSouth BeLafayettSimi Val
Springfi
InglewooPeoria cIndependWaco citBeaumontAnn ArboAbilene
El Monte
ColumbiaVallejo ThousandStamfordFort ColLansing Elizabet
Cedar RaFayettevHartfordEvansvilConcord Topeka cNew HaveSioux FaSterlingMesquite
Flint ciFullerto
Alexandr
Orange c
SavannahSunnyvalEscondidPasadenaSalem ciEugene cTorranceWarren cHollywooBridgepoBrownsviHayward PasadenaAurora cLakewoodHampton Kansas CSyracuseSanta RoOverland
PatersonPomona c
RockfordTallahas
Salinas
SpringfiSpringfi
Fort LauChattano
Ontario
HuntsvilTempe ciOceansidGarden G
Dayton c
Oxnard c
WorcesteChula ViProvidenAmarilloKnoxvill
Laredo c
Newport Reno citSalt LakLittle RJackson San BernWinston-Boise CiOrlando
ColumbusDurham cModesto HuntingtIrving cTacoma cGlendaleSpokane Yonkers RichmondGrand RaDes MoinMobile cChesapeaLubbock ShrevepoMontgomeScottsda
Fremont
Fort WayMadison Garland Akron ciGlendaleRochesteGreensboLincoln
Hialeah
Baton RoNorfolk Jersey CBirmingh
StocktonBakersfi
St. Pete
Riversid
LouisvilLexingtoNewark cRaleigh Aurora cCorpus CSt. Paul
Buffalo Tampa ciToledo c
Anaheim
CincinnaArlingtoPittsbur
Santa An
Wichita St. LouiColoradoMiami ciHonoluluMinneapoOmaha ciTulsa ciMesa cit
Oakland SacramenAtlanta VirginiaFresno c
Kansas CAlbuquerLong BeaClevelanLas VegaNew OrleTucson cOklahomaPortland
Fort WorCharlottDenver cSeattle
El Paso Nashvill
WashingtBoston cMilwaukeMemphis BaltimorAustin cColumbusJacksonv
San FranIndianapSan Jose
Detroit San Anto
Dallas cSan DiegPhoenix
PhiladelHouston
Chicago Los Ange
New York
And in changesLo
g C
han
ge
Ho
usin
g U
nits
, 19
70-2
000
Figure 3: Log Changes in Housing Units and Population, 1970-2000Log Change Population, 1970-2000
log change tot housing units, 1 Fitted values
-.580574 1.84035
-.344024
2.1741
St. Loui
JohnstowGary citYoungsto
Detroit
Buffalo ClevelanPittsburFlint ciNiagara WheelingUtica ciSaginaw
Dayton cHuntingtLouisvil
Newark c
Baltimor
HarrisbuCincinnaCanton cScrantonBinghamtRochesteSyracuseCharlestParkersbLima citWashingtNorfolk
Hartford
AnnistonWarren c
PhiladelAltoona Akron ciRichmondMacon ciJamestowJackson
Erie citBirminghWilliamsYork cit
New Lond
Trenton New OrleToledo cAlbany cMilwauke
Atlanta Anderson
Atlantic
Terre HaSouth BeRacine c
Chicago Duluth cKansas CKansas CLorain cEvansvil
Galvesto
BerkeleyMinneapoPeoria cBridgepo
MansfielHamiltonNew HaveMuskegon
KalamazoFitchburDecatur WilmingtLansing PortsmouWaterloo
Livonia Greenvil
Boston c
Lake Cha
Jersey CNew BedfSt. PaulReading SpringfiPoughkeeMonroe cPensacolPawtuckeFall RivYonkers Providen
Allentow
Muncie cWorcesteLancaste
Topeka cBeaumontPortland
Sioux CiDes MoinWaterburKnoxvill
DavenporGrand RaNew York
IndependRockfordHagersto
Columbia
Torrance
Paterson
Roanoke
Salt Lak
Memphis Danville
Biloxi c
Mobile cNorwalk Pueblo cHouma ciBremerto
Brockton
Albany c
Seattle
Florence
Elizabet
Stamford
Lawrence
Denver cMiami ciWichita
San Fran
Cedar RaFort LauTampa ciShrevepo
Oakland
Savannah
Lowell cAlexandrOmaha ciState Co
TexarkanAnn ArboHonoluluKenosha Spokane
Huntsvil
St. Pete
AlexandrFort Way
CharlottJoplin c
Odessa c
New Brun
Middleto
Wausau c
Pasadena
Tuscaloo
Tulsa ciElkhart
Champaig
Waco cit
AshevillJackson Madison LynchburHampton SpringfiManchestAppletonBurlingt
Lawton cWichita
Inglewoo
Tacoma c
LafayettSpringfiFort SmiJanesvil
Long Bea
Abilene ChattanoHollywoo
Newport
Sarasota
Los AngeSanta Ba
ColumbusLubbock
Garden G
Joliet c
Corpus CFort WorAmarilloBattle C
Baton RoOklahoma
Little REau ClaiSunnyvalSan AngePortland
Winston-
Dallas cDaytona Boulder Decatur Concord West Pal
ColumbiaCharlestTyler ciBillings
Fullerto
Glendale
Danbury
Gainesvi
Hayward
Lincoln Montgome
Santa FeGreensboLakewoodDothan c
Yakima c
Houston PasadenaMidland Rocheste
Lafayett
SacramenLongview
Blooming
BloomingHuntingt
Johnson Wilmingt
El Monte
Orange c
Fargo ci
Santa Cr
Bellingh
Pomona c
Sioux FaEl Paso San AntoSan DiegVallejo
Melbourn
San Bern
Eugene c
Riversid
AlbuquerTucson cOrlando Fayettev
Aurora c
Overland
Bryan ci
Durham cLas CrucSimi ValIrving c
Anaheim
Greeley Provo ci
San JoseSalem ci
Fremont
Tallahas
Santa An
Hialeah
ChesapeaCharlottMesquiteFayettev
Stockton
Phoenix
Raleigh
Santa MaOxnard c
Killeen
Ontario
Virginia
Reno citBoise Ci
Tempe ci
Chula ViLaredo c
Salinas
Fresno c
Austin cGarland BrownsviColorado
Fort ColMcAllen
Santa Ro
Scottsda
Modesto
ThousandClarksviBakersfi
Escondid
ArlingtoAurora cLas Vega
Oceansid
Glendale
Mesa cit
Vacancy Rate Coefficient is .1C
ha
ng
e in
Va
cancy
Ra
te, 19
90
-2000
Figure 4: Changes in Population and Vacancy Rates, 1990-2000log change population, 1990-2000
abs change in vacancy rate, 199 Fitted values
-.162849 .616416
-7.47307
7.59984
Johnstow
Youngsto
Hartford
St. Loui
Lima citGary cit
Utica ciBaltimor
Flint ci
Saginaw
Buffalo
Binghamt
Norfolk
SyracuseNiagara
New Lond
Wheeling
PittsburCincinna
Anniston
Macon ciDanville
Dayton c
Birmingh
Jamestow
Galvesto
Detroit
Charlest
Scranton
Harrisbu
Lansing Jackson New Bedf
Huntingt
Toledo c
AlexandrWashingt
ClevelanAlbany c
New Have
Rocheste
Fitchbur
Muncie c
Milwauke
Louisvil
Erie cit
Warren c
Altoona Savannah
PhiladelCanton c
Greenvil
Kalamazo
Williams
Evansvil
Trenton
Lorain c
Pensacol
Monroe c
Portsmou
York cit
SpringfiJackson
Racine c
Akron ci
Richmond
Bremerto
Mansfiel
New Orle
Decatur ParkersbKansas C
Waterbur
Roanoke
Bridgepo
Albany c
State Co
Lynchbur
Hamilton
Huntsvil
Fall Riv
Florence
Newark c
Peoria cMuskegon
Beaumont
Battle C
Livonia
Portland
Berkeley
Tuscaloo
Pawtucke
Anderson
Shrevepo
Independ
Middleto
Miami ci
Odessa c
Mobile c
Lancaste
Allentow
Kansas C
Wilmingt
Brockton
Duluth c
Lake Cha
Lowell c
Worceste
Honolulu
PasadenaChattano
Fort Lau
Topeka c
South Be
Boston c
Lawrence
Inglewoo
Des Moin
DavenporSarasotaWaterloo
Hagersto
Pueblo cDaytona Poughkee
Reading
Torrance
Wausau c
Terre Ha
Baton Ro
MinneapoSt. Pete
Chicago Ann Arbo
Little R
Yonkers
LongviewHuntingt
Grand Ra
San Ange
Jersey C
Knoxvill
St. Paul
Sioux Ci
Atlanta
Norwalk
Paterson
Springfi
Newport Los Ange
Midland
Houma ci
Champaig
Memphis
Atlantic
Appleton
Vallejo
Tulsa ci
Lubbock
San FranOakland
Manchest
Long Bea
RockfordMontgome
Dothan c
Corpus C
SpringfiSanta Ba
ProvidenVirginiaWichita Glendale
Stamford
Tampa ci
Eau Clai
Abilene
Madison
Seattle El Monte
Biloxi cNew York
Concord
El Paso
Hampton Tacoma c
Elizabet
Waco cit
Texarkan
Amarillo
San Dieg
Sacramen
Fort SmiFullerto
Spokane Decatur
BillingsTyler ci
Cedar Ra
Santa Fe
Joplin c
Simi ValSanta Cr
CharlottAshevill
Tempe ci
Thousand
Sunnyval
ColumbusKenosha
GainesviRiversid
Johnson
Orlando
San Bern
Wichita
Salt Lak
Boulder
Oklahoma
Lakewood
Burlingt
Danbury
Janesvil
Blooming
San Jose
Hollywoo
Modesto
Santa AnLawton c
Alexandr
Garden GStocktonOmaha ciOrange c
New BrunAlbuquer
Lafayett
Fremont
Lincoln
Dallas c
Columbia
Denver c
Ontario
Pasadena
Fort WayElkhart
Bryan ci
Garland Fort Wor
Las Cruc
Melbourn
Houston
Oxnard c
Tucson c
CharlestHialeah
Tallahas
Fresno cPortlandProvo ci
Rocheste
West Pal
GreensboFargo ci
San Anto
Columbia
Eugene c
MesquiteEscondid
Sioux Fa
Anaheim
Irving c
Aurora c
Blooming
Oceansid
Hayward Santa Ma
McAllen
Salem ci
Greeley
ArlingtoColorado
Chula Vi
Bellingh
Lafayett
Winston-Santa Ro
Yakima c
Chesapea
Raleigh
Overland
Phoenix
Reno cit
Fort Col
WilmingtCharlott
Killeen
Durham cClarksvi
Mesa cit
Fayettev
Joliet c
Salinas
Austin c
Brownsvi
BakersfiAurora c
Laredo c
Glendale
Boise Ci
Scottsda
Fayettev
Las Vega
Durable Housing: the Basic Idea
Durable housing is needed to explain American Cities
Me
dia
n ho
use
pric
e, $
200
0
Figure 2: Median Price Regression and Construction CostsFitted values
medval8000 Fitted values Construction Costs
18799.4 192187
18799.4
218594
Newark c
Buffalo
Providen
Dayton cSt. Loui
Atlanta ClevelanSyracuse
Boston cNorfolk
Detroit
SpringfiJersey C
RochesteLouisvilBaltimor
Philadel
Knoxvill
Cincinna
Birmingh
New York
Pittsbur
Spokane
Worceste
MinneapoSalt Lak
Akron ci
Chicago
El Paso RichmondGrand RaColumbusChattano
New Orle
Memphis
Kansas C
ColumbusMilwauke
St. Paul
Miami ci
Gary citSan Anto
Fort Way
Colorado
Flint ci
Long Bea
Des Moin
Omaha ciToledo c
Tampa ci
Denver c
St. Pete
Fresno c
Washingt
Kansas C
Jackson Tacoma c
Lexingto
Fort WorIndianap
Oakland
Tucson cLincoln
MontgomeNashvill
OklahomaLubbock ShrevepoMobile cGreensbo
PortlandAlbuquer
Little R
Jacksonv
Wichita
Madison Sacramen
Tulsa ciDallas c
Baton Ro
Austin c
Yonkers
Corpus C
Las Vega
Raleigh
Los Ange
VirginiaMesa cit
Phoenix
San Fran
Seattle
Santa An
San Dieg
Aurora c
Fort Lau
Riversid
Houston
Arlingto
Anaheim
San Jose
Implications of a Durable Housing Model
1. Population growth rates will be skewed because places grow quick and decline slowly.
2. There will be strong persistence of growth rates especially in decline
3. Places with housing costing below construction costs will not grow
4. Positive shocks increase population more than housing prices; negative shocks decrease housing prices more than population
5. Concave correlation between prices and growth
Results on Durable Housing
• Highly skewed distribution of growth rates• Coefficient of current growth on past growth is 1
if growth was negative and .4 if growth was positive
• Coefficient on price growth on population growth is 1.8 when negative and .2 when positive
• Strong relationship between housing below construction cost and no population growth
The Concave Price/Growth Relationship
Pric
e A
ppre
cia
tion,
197
0-2
000;
1=
100%
Figure 3: Price Appreciation and Urban GrowthPopulation Growth, 1970-2000
real median house price appreci Fitted values
-.44 0 1 2 3 4 5
-.28
0
1
2
3
4
St. Loui
JohnstowYoungstoGary citDetroit Buffalo ClevelanPittsbur
Flint ciNiagara Wheeling
Utica ciSaginaw
Dayton cHuntingt
LouisvilNewark cBaltimor
HarrisbuCincinnaCanton c
Binghamt
Scranton
RochesteLima citSyracuseParkersb
Washingt
Norfolk
Hartford
Anniston
Warren cPhiladel
Altoona
Akron ciRichmondBirminghMacon ci
Jamestow
CharlestJackson Erie citWilliamsNew Lond
York cit
Trenton
New OrleToledo cKansas CAlbany cMilwaukeMuncie c
Atlanta
AndersonSouth Be
Atlantic
Terre Ha
Mansfiel
Chicago
Racine cDuluth cKansas CEvansvilGalvestoLorain c
Berkeley
Minneapo
Peoria cBridgepoHamilton
New Have
Roanoke
MuskegonKalamazoFitchbur
Wilmingt
Decatur
Portsmou
Lansing Waterloo
Greenvil
Livonia
Boston c
Lake Cha
New Bedf
Jersey C
St. PaulReading SpringfiSpringfiPoughkeeAlexandrMonroe c
PensacolPawtucke
Fall Riv
Yonkers
BeaumontProvidenAllentowLancasteWorceste
Topeka cDes Moin
Portland
Sioux CiMemphis Waterbur
KnoxvillIndianap
Davenpor
Grand RaLynchbur
New York
IndependRockford
Pueblo cHagersto
Columbia
Torrance
Paterson
Salt Lak
BremertoDanville
Biloxi c
Norwalk
Mobile cHouma ci
Brockton
Albany c
Seattle
TexarkanFlorence
Elizabet
Stamford
Lawrence
Denver c
Wichita
Miami ci
San Fran
Omaha ci
Cedar Ra
Fort LauTampa ci
Shrevepo
Oakland
Savannah
Fort WayColumbus
Lowell cAshevill
Tuscaloo
Jackson Huntsvil
Odessa c
Ann Arbo
Honolulu
Kenosha
Spokane
St. Pete
Champaig
Alexandr
CharlottJoplin cNew BrunMiddletoState Co
Wausau c
Burlingt
Pasadena
Tulsa ciWaco citElkhart
Madison
Hampton
Manchest
AppletonWichita Lawton c
Inglewoo
Tacoma c
Lafayett
Nashvill
Fort SmiJanesvil
Long Bea
Abilene
Chattano
Newport HollywooSarasota
Los Ange
Santa Ba
ColumbusLubbock
Joliet c
Garden G
Charlest
Corpus CFort WorAmarilloBattle CBaton RoOklahoma
Little REau Clai
Portland
Sunnyval
San AngeWinston-Jacksonv
Tyler ciDallas cDaytona
Boulder
Decatur
Johnson
Billings
BloomingWest Pal
Concord
ColumbiaMontgome
Fullerto
Glendale
Danbury Lincoln
Gainesvi
Yakima c
Lexingto
Hayward Santa Fe
Greensbo
Longview
Lakewood
Houston Dothan c
Sacramen
PasadenaRochesteMidland
LafayettSan Anto
Blooming
Vallejo
HuntingtWilmingt
Springfi
El Monte
Orange c
Fargo ci
Santa Cr
Bellingh
Pomona c
Sioux Fa
Eugene c
El Paso
San Dieg
MelbournSan BernLakelandTucson c
RiversidAlbuquer
Orlando Fayettev
Salem ci
Aurora cOverland
San Jose
Bryan ci
Durham c
Las Cruc
Simi Val
Irving c
Anaheim
Greeley Provo ci
Fremont
SterlingTallahas
Fayettev
Santa AnBoise Ci
Hialeah Stockton
ChesapeaCharlott
Phoenix
Raleigh
Mesquite
Santa Ma
Clarksvi
Oxnard c
Killeen
Ontario
VirginiaReno citTempe ciBrownsviFresno c
Laredo c
Chula Vi
Salinas
Austin cColorado
Garland
Fort Col
McAllen
Santa Ro
Modesto
Scottsda
Thousand
Bakersfi
Escondid
Arlingto
Aurora c
Las Vega
Oceansid
GlendaleMesa cit
The Weather and Urban Growth
• We split the weather into positive shocks and negative shocks so that the same share is negative as had overall population declines
• The coefficient on weather and price growth is .006 for negative shocks and .002 for positive
• The coefficient on weather and population growth is .0008 for negative shocks and .068 for positive shocks
A Final Implication: Durable Housing and Poverty
• If cities decline by becoming less productive, and if productivity relates to skill level
• Then poor people will stay in declining cities disproportionately because they have cheap, durable housing
• Poor people do congregate in declining cities, but this disappears when you control for housing prices
The Regulatory Tax
• Housing Supply Costs, in growing areas, are C*S+R where C is structural cost, S is size of structure and R is residual
• In most of U.S. history, the 1970, R/(CS+R) is small– less than .2 almost everywhere
• Only over the past thirty years do prices start to greatly exceed construction costs
Why the gap between housing prices and construction costs?
• Theory # 1: Land is expensive
• Theory # 2: Regulation prevents new construction
• R=PL+T where P is land costs, L is land area and T is regulatory tax
• We don’t directly observe land costs, but we can estimate them hedonically
• R/L>10*the estimate of P– tax not land
Another piece of Evidence: NYC
• In New York City, apartments are always the cost of building up
• No matter what the fixed costs are, the marginal cost is technological and generally less than 200$/square foot
• Yet condo prices are now often over 600$/square foot
• Hard to reconcile with a free market
Other Evidence on Rising Regulation
• Declining numbers of permits
• Little correlation between prices and density across metro areas
• Correlation between changes in prices and changes in population has become negative across regions
• Places with more estimated “zoning tax” have other measures of regulation
The Change in the Price/Quantity Relationship
In NYC in the 50s and 60s, rising prices related strongly to new permits.
In the 80s and 90s, this positive relationship has disappear.
Anecdotal information strongly supports the idea that citizens groups can now block change, presumably to keep prices up.
We don’t know why this occurred.
Supply Restrictions and Urban Dynamics
• Any restrictions on new supply will change the way that cities develop.
• One possible source of restricted supply is zoning, but limited land is certainly another.
• This will change the ways cities develop– compare Massachusetts and Texas
Supply and Urban Growth
Price
Number of Homes
TexasSupply
MA Supply
Rise in Demand
Massachusetts Population
Figure 13: Price Growth and Schooling in TexasShare w /College Degrees 1980
Housing Price Grow th .
0 20 40 60
.329786
1.0161
Port Art
Laredo c
Pasadena
Brownsvi
Mesquite
Grand Pr
Harlinge
San Anto
Victoria
Odessa c
El Paso
Baytown
Corpus C
San Ange
Wichita Killeen
Amarillo
Waco cit
McAllen Longview
Fort Wor
BeaumontIrving cGalvesto
Temple c
North RiAbilene
Tyler ci
Garland
Dallas c
Lubbock
Houston
Bryan ci
Midland
Arlingto
Carrollt
Austin c
Denton c
Plano ci
Richards
College
Figure 13: Price Growth and Schooling in TexasShare w /College Degrees 1980
Housing Price Grow th .
0 20 40 60
.329786
1.0161
Port Art
Laredo c
Pasadena
Brownsvi
Mesquite
Grand Pr
Harlinge
San Anto
Victoria
Odessa c
El Paso
Baytown
Corpus C
San Ange
Wichita Killeen
Amarillo
Waco cit
McAllen Longview
Fort Wor
BeaumontIrving cGalvesto
Temple c
North RiAbilene
Tyler ci
Garland
Dallas c
Lubbock
Houston
Bryan ci
Midland
Arlingto
Carrollt
Austin c
Denton c
Plano ci
Richards
College
Massachusetts Prices
Figure 15: Price Growth and Schooling in Mass.Share w /College Degrees 1980
Housing Price Grow th .
0 20 40 60
1.00418
1.8703
New Bedf
Fall Riv
Lawrence
Taunton
Chicopee
Lynn cit
BrocktonLowell c
HaverhilMalden c
Springfi
Medford
Worceste
Quincy c
Somervil
Waltham
Boston c
Cambridg
Newton c
Texas Prices
Figure 13: Price Growth and Schooling in TexasShare w /College Degrees 1980
Housing Price Grow th .
0 20 40 60
.329786
1.0161
Port Art
Laredo c
Pasadena
Brownsvi
Mesquite
Grand Pr
Harlinge
San Anto
Victoria
Odessa c
El Paso
Baytown
Corpus C
San Ange
Wichita Killeen
Amarillo
Waco cit
McAllen Longview
Fort Wor
BeaumontIrving cGalvesto
Temple c
North RiAbilene
Tyler ci
Garland
Dallas c
Lubbock
Houston
Bryan ci
Midland
Arlingto
Carrollt
Austin c
Denton c
Plano ci
Richards
College
Texas Population
Figure 12: City Growth and Schooling in TexasShare w /College Degrees 1980
Population Grow th .
0 20 40 60
-.119038
1.12156
Haltom C
Port Art
Laredo c
Texas Ci
Pasadena
Brownsvi
Mesquite
Del Rio
Grand Pr
Harlinge
Texarkan
San Anto
Victoria
Odessa c
El Paso
Baytown Corpus CSan Ange
Wichita
Killeen
AmarilloLufkin cWaco cit
McAllen
Longview
Fort Wor
Beaumont
Sherman
Irving c
Galvesto
Temple c
North Ri
Kingsvil
Abilene Tyler ci
Garland
Hurst ci
Dallas cDuncanvi
Lubbock
Houston
Nacogdoc
Bryan ci
Midland
Arlingto
Carrollt
Austin c
Denton c
Plano ci
Richards
College
Responses to Labor Demand Shocks
Regulation (Wharton Survey)
Change in Population
Change in Income
Change in Prices
Low Regulation
1.04
(.3)
11597
(3917)
54899
(37478)
High Regulation
.2
(.3)
34651
(13007)
204730
(137972)
The Costs of Rent Control
• Undersupply• Reduced maintenance• Social waste on rent seeking• Misallocation (Deacon and Sonstelie,
Hubert, Suen)• Nat Sherman rented a six month CPW
apartment for 335/month and said the apartment “happens to be used so little that I think [the rent is] fair”
Misallocation under Rent Control
Rent
Transfer
SurplusLeft
DWL
Quantity
Demand
Supply
Misallocation under Rent Control
Rent
Transfer
SurplusLeft DWL
Quantity
a
b
c
ab=bc
MisallocationLoss
Demand
Supply
How Big is the Misallocation Loss?
• The misallocation loss is technically first order while the undersupply loss is second order
• Thus for sufficiently mild impositions of rent control the social loss is always greater from misallocation
• This relies on random matching– better matching would reduce losses
• Different impacts of demand elasticity --
Empirical Approach
• Assume that if household A consumes more of attribute y than household B in city 1, this will also be true in city 2.– This assumes the we can rank households by
demand.
• For any city c and subgroup i, the distribution of demand, f(d, x) equals f(d+lc, x) for some lc.– This assumes the all demand shifts are city
specific
These assumptions imply Constant Overlap
• If the share of subgroup i in the free market city A that rents apartments with k or fewer rooms is equal to the share of subgroup i in free market city B that rents apartments with n or fewer rooms, then for any other subgroup j, the share renting apartments with k or fewer rooms in city A must equal the share renting apartments with n or fewer rooms in city B.
Results
• In NYC, 47 percent of high school dropouts consume more rooms than people with college degrees (31.6 percent for the U.S. as a whole)
• In NYC, 45.7 percent of people in the bottom third of the income distribution consume more rooms than people in the top third– in the U.S. the number is 35.1 percent.
Full Structural Estimation
• Estimate the maximum cutoff of unobserved demand within each demographic group associated with each apartment size
• Calculate the total amount of misallocation: 20.9 % with correction for sampling error
• By comparison renters in Hartford (4 percent), Chicago, 7 percent
• Misallocation is strongest in Manhattan (26 percent) and among long term residents
Homeownership and Social Capital
• What is social capital? • One view is that it is socially-relevant human
capital that is determined by investment
– Social characteristics, including charisma, status and access to networks, that enable that person to extract private returns from interactions with others
• Social capital can be individual or aggregated up to form society-wide social capital
Is so, then usual investment models can understand this thing?
• Social capital should rise and fall over the lifecycle (it seems to)
• People in more social occupations should invest more (they do)
• People who are more patient or just invest more generally should invest more in social capital (they do using education)
• People who are more mobile will invest less
Homeownership and Social Capital
• Homeowners have more expected permanence and have a property stake in the quality of the community
• They should invest more in local public goods, at least that is one of the stated reasons for subsidizing ownership
• But how big are these effects really?
• And are the subsidies effective?
Are Homeowners Better Citizens?
• Using almost all measures of social capital, people who are homeowners are better citizens– .25 more organization, .09 knows school head, .10
knows US representative, .15 votes in local elections– Also, .12 garmed and .1 owns a gun– ½ of the good effects are related to permanence
• These effects however are much bigger without controls, because homeowners are really different based on observables
• The selection problem is huge
The Endogeneity Problem
• Unsolved but two approaches– first use area averages based on structures– same basic results
• Second use GSOEP data from Germany where you have a panel– Much smaller impacts in general– With fixed effects home repair drops from .12 to .09– Volunteering drops from .033 to .013 and poltiical
participation from .04 to .008– Effects are small but significant statistically
But does the subsidy do anything?
• Homeownership is essentially determined by structure– 85 percent of people in houses are owners, 85 percent of people in apartments are renters– Incentive Problems
• Homeownership doesn’t change much over time at all even though subsidy changes with inflation
• Across people are well, the size of the subsidy doesn’t seem to matter
Cities and Governments
• National governments play a huge role in shaping cities– Large scale infrastructure spending– Place based initiatives and redistribution– Transport technologies
• Local governments are also critical– Schools, Safety, Other Services– Local Redistribution
Trade and Circuses: Mega-Cities
• What determines the level of primacy across countries?
• Krugman and Livas point to international trade– because trade is space neutral (is it?) the incentive agglomerate declines
• High internal transport costs is presumably another reason to agglomerate
The Political Roots of Agglomeration
• A dictator’s desire to invest may decline with distance from him – Investment for consumption reasons– Investment to deter unrest
• Political influence declines with political distance– Physical threat declines– Lobbying, etc., also declines
• This should be more important in unstable or dictatorial regimes
Capital Cities and Transfers
• As a result, capital cities have generally received more benefits from government
• Sometimes these reflect dictators building themselves nice cities (St. Petersburg)
• Sometimes it is a response to political power of locals (Washington)
• Sometimes it is a response to local uprisings (students in Santiago)
The Empirical Causes of Mega-Cities
• Basic specification • Log(Primate City Population)=• a*Log(Non-Urban Population)• +b*Log(Urban Population)+• Country Factors • Countries with higher levels of trade do indeed
have smaller central cities (-.6)• Internal investment in roads matters (but what
about causality)
Politics is very significant
Stable Democracies
Primacy=.23 (.03)
N=24
Stable Dictatorships
Primacy=.3 (.03)
N=16
Unstable Democracies
Primacy=.35(.07)
N=6
Unstable Dictatorships
Primacy=.37 (.02)
N=39
In regressions:
• Capital City Effect = .42 (only 8 non-capitals)
• Dictatorship Effect= .44 (.15)
• Dictatorship + Instability + Interaction yields .7, 2.3 and -2.3 all significant
• But does primacy lead to dictatorship or dictatorship to primacy
Tests for Causality
• Instrument using various political variables such as ethnic heterogeneity (predates the cities) and neighboring instability-- .5
• Between 1970 and 1986, dictatorships in 1970 had faster growth in the primate city
• However, there is no significant relationship between size of capital city and becoming a dictatorship
History
• Rome’s growth peaked between 135 and 50 b.c.e. when it grew from 375k to 1,000,000.
• Strength abroad and weakness at home leades to redistribution to the capital
• Empire expanded in Gaul, Bithynia, Pontus, Cilicia and Syria
• Pompey declares all conquests are part of the city governemnt
• Sempronian and Clodian laws extrend the grain distribution to Italians in Rome
• Sulla extends citizenship to all inhabitants of Italy
Other Cities
• Edo (Tokyo) expands from little to between 500k + 1 million in the 1600s
• Growth entirely related to being Shogunal capital for newly unified Japan
• Buenos Aires grew most between 1870 and 1914– industry and politics (London)
• Paris and Mexico city are more overtly political
Local Governments
• There is a strong traditional from Tiebout (and the Federalist papers) that suggests that many benefits of local governments
• Opportunities for variety• An ability to influence outcomes through
both voice and exit• However, local governments are
particularly bad at redistribution because of mobility
Incentives and Local Governments (Public Choice, 1996)
• Inducing Local Governments to behave well presumably requires incentives
• Taxes can provide those incentives is governments want more revenue
• Property taxes have the benefit of inducing long time horizons for governments
• Tradeoff between income and property taxes involves the elasticity of demand for space (highly elastic– income tax looks better– inelastic– housing is better)
General Redistribution Point
• If the average tax rate (pure redistribution) is determined by the income level of the median voter t(y) and
• The income level of the median voter is determined by the level of redistribution y(t)
• Then the local equilibrium is determined by the point where y(t(x))=x
Median Income in the City
Tax
Rat
e fo
r R
edis
trib
utio
n
Tax Rate as a function of Income
Income as aFunction of Tax rate
The Curley Effect
• Tiebout suggests that since localities will want their communities to grow, this will create good incentives for governments
• But what if governments don’t want their cities to grow (as in zoning)
• Or even worse, what if they want their cities to lose their richest residents
Shaping the Electorate
• James Michael Curley was the mayor of Boston on four separate occasions from before WWI to after WWII
• He was highly focused on ethnic conflict and also ended up in jail
• When asked in WWI, if a UK recruiter could recruit Bostonians of British extraction to fight, Curley replied: “Go Ahead, Take Every Damn One of Them”
Curley’s Logic
• The rich anglo-bostonians were never going to vote for him
• As a result, by eliminating them he increased his vote share
• This requires some form of group identification
• This can also be seen in the policies of African-American mayors like Berry or Coleman Young (Detroit)